The electric power grid is a complex system under semi-autonomous distributed control. However, visibility into the distribution grid is limited. While the distribution grid is relatively reliable, anomalies and outages occur, which can be related to factors such as weather and stochastic generation, equipment degradation, or physical and cyber security attacks. These anomalies and operational characteristics can have significant economic and physical impact (e.g., damaged equipment) to both consumers and utilities. There is a need to develop reliable means to anticipate and recover from anomalies and outages, while also enabling a high penetration of renewables in the future.
The project goal is to develop advanced supervisory and distributed algorithms for DER and microgrid coordination that rely on uPMU measurements, other grid sensors and prior knowledge on mobility patterns of electric vehicles. We are investigating different coordination schemes among multiple microgrids, aiming to alleviate distribution system problems such as over/under voltage and frequency and congestion.
This project is focusing on applying signal processing, sensor fusion, non-linear optimization techniques, and machine learning techniques to:
- Detect and classify anomalies and regular operation in the power distribution grid
- Understand the variability in the availability of distributed energy resources
- Devise control strategies that are robust to such variabilities in order to mitigate anomalies and coordinated Distributed Energy Resources (DER) and microgrids.
Development of these coordinated control algorithms are challenging for several reasons. First, since microgrids are individually owned, the coordinated control must be fair in economic burdens (including cost of lost opportunity) imposed on various microgrids. Second, microgrids may not allow the control algorithm access to the current operating status of their resources for security and privacy purposes, and hence, the control algorithm may need to act solely based on the uPMU data. Third, consider models of shared resources (EVs). Finally, the control algorithm must operate in several time frames since the problems that need to be addressed occur in different time frames, e.g. reactive power and voltage problems in the system happen in much slower time frames compared to over/under generation problems.
Furthermore, we will achieve a framework for identification of grid anomalies with distribution PMU and traditional utility data. As well as understand the inherent tie between transportation needs and impacts on power distribution system through systematic analysis of complementary and novel data sources in the field.
Related Links: Laboratory Directed Research and Development (LDRD)
Project Partners: Computer Science Devision